- within Intellectual Property topic(s)
- within International Law and Compliance topic(s)
- in Turkey
1. Copyright Protection: Automatic Protection upon Creation
Under Turkish law, software is protected within the scope of the Law on Intellectual and Artistic Works No. 5846 ("Copyright Law") and is recognized as a "literary and scientific work". For software to qualify as a "work" and benefit from copyright protection, it must (i) be the product of intellectual effort, (ii) reflect the creator's individuality, and (iii) be expressed in a tangible form.
Copyright protects the form of expression of software, meaning the embodied form of the software such as codes, not the idea or functionality. Algorithms, technical ideas, or the functions provided by the software remain outside the scope of this protection. Copyright protection arises automatically upon the creation of the software, and no registration or notification requirement is imposed.
The most significant advantage of copyright protection is that protection begins as soon as the software is physically created. However, this protection covers only the code itself; it does not protect the technical solution or functionality provided by the software. This situation creates a significant gap, particularly in competitive technology markets, preventing competitors from copying the code but not preventing the functionality of the software from being implemented using a different coding method.
2. Patent Protection: The Shield for Technical Innovation
Both the Turkish Industrial Property Law ("IP Law") and the European Patent Convention ("EPC") exclude "computer programs as such" from patentable subject matter. However, when software demonstrates technical character (for example, by producing a tangible technical effect), it may benefit from patent protection, provided it also meets the criteria of novelty, inventive step, and industrial applicability.
Although the concept of "technical character" is not explicitly defined by legislation, it plays a decisive role in patentability assessments. According to the European Patent Office ("EPO") approach, for software to be accepted as having technical character, it must produce results beyond the normal physical interactions between the program and computer hardware.1
There are two different approaches in determining technical character: "further technical effect" and "any hardware". According to the "further technical effect" approach, for software to be accepted as having technical character, there must be a technical effect that goes beyond the normal physical interactions between the software and the computer. Whether a further technical effect exists will be determined according to the specifics of each concrete case.2 For the emergence of a further technical effect to be accepted, it is irrelevant whether this effect occurs inside or outside the computer. Accordingly, even if the further technical effect emerges inside the computer as a result of software running on the computer, it can be protected by patent.
For instance, whilst a program that calculates tax values would not be considered technical, an application that enables a phone's camera to recognize objects (image processing and object recognition algorithms) or software that optimizes machine performance can be accepted as having technical effect.
The "any hardware" approach, in addition to the further technical effect, also requires the presence of hardware (such as a computer or smartphone) for patentability. In practice, both the EPO and the Turkish Patent and Trademark Office ("TPTO") follow the "any hardware" approach when examining software-related inventions. In this context, when technical devices such as computers or mobile phones are specified in the claims, the invention is accepted as having technical character, and the evaluation proceeds to the criteria of novelty, inventive step, and industrial applicability. 3
The most debated topic in recent years, particularly within the scope of patent protection for software, is the position of artificial intelligence ("AI") within patent protection. As with traditional software, AI-related inventions can benefit from patent protection if they exhibit technical character and satisfy the general patentability requirements.
On the other hand, it is not always possible for AI models, just like traditional software, to produce a technical result. For example, whilst AI systems that merely generate text, classify documents, or perform cognitive-like tasks without a technical purpose are deemed non-patentable, AI models applied in technical contexts, such as detecting irregular heartbeats in a medical device, can benefit from patent protection if claimed together with a physical element under the "any hardware approach".
When novelty lies primarily in the data set used for training rather than the structure or functioning of the AI model, patentability becomes complex. This is because data, as such, is not a technical contribution and is generally suitable for protection under trade secret provisions.4 In contrast, where innovation lies in the AI model's architecture or structural components, for example, if the model is completely newly developed or technical modifications and improvements are made to the structure of an existing model, these structural innovations can form the basis of a patent application.4
The question of whether an invention is patentable when the novelty element is concentrated in the data set is the focal point of the U.S. Court of Appeals for the Federal Circuit's decision in Recentive Analytics Inc. v. Fox Corp., dated April 18, 2025. The Court emphasized that merely applying established machine learning methods to a new data environment is not patentable. Innovation cannot consist merely of using a different data set; there must be an original contribution in the model's architecture, training method, or technical application. 5
3. Trade Secret: A Solution to the Black Box Problem
In a patent application, the technical problem the invention solves and how this problem is solved must be clearly and completely disclosed.6 For example, if the element contributing to the invention is an algorithm, this algorithm must be clearly presented. However, if the novelty relates only to the use of data and the algorithm is not part of the invention, disclosure of the algorithm may not be mandatory.
AI inventions raise practical challenges in satisfying the patent system's disclosure requirement—the obligation to clearly and completely describe the invention so that it can be reproduced by others. Many AI systems function as "black boxes"; even developers cannot fully explain how the system reaches its results. Explaining step-by-step the decision-making process of a neural network with millions of parameters can be practically impossible.
Global intellectual property offices, including the IP5 (U.S., Japanese, European, Korean, and Chinese patent offices), have recently emphasized that AI-related inventions must still be disclosed in a manner that is reproducible and repeatable.7
In our country, trade secrets are protected within the scope of the unfair competition provisions of the Turkish Commercial Law No. 6102 ("TCL") and certain special laws. Considering that patent protection is provided for a maximum of 20 years under the IP Law in balance with general public interests, that registration processes are lengthy and costly, that developments in the technology field are very rapid, and that the requirement to make technical disclosures during the registration process facilitates competitors' access to technical information, it is clear that patent protection is not always in the interest of the right holder. In such cases, protecting software and technical solutions as "trade secrets" may be more suitable for the right holder's interests.8
4. Conclusion and Strategic Recommendations
In today's technology-driven environment, software and AI systems blur the lines between artistic expression and technical invention. Combining intellectual property tools smartly and adopting a hybrid protection strategy is critically important.
Determining which tools can protect software and AI-based systems in intellectual property law requires a multi-layered evaluation from both technical and legal perspectives. In light of Turkish law and international practices, it is clear that software can be subject to copyright protection in terms of its form of expression; however, it is also possible for software containing technical contributions and meeting certain conditions to be protected by patents.
Specifically for AI, factors such as the existence of technical contribution, the feasibility of meeting the disclosure obligation, and whether the novelty is reduced solely to data use are determinative in the patentability assessment. Nevertheless, the choice between trade secret protection and patent protection, given AI's "black box" nature, is also critically important.
The key lies in identifying where the innovation truly resides—whether in creative code, technical architecture, or intelligent data use—and tailoring an intellectual property strategy that secures both legal protection and business advantage.
Footnotes
1.EPO, Guidelines for Examination in EPO (01.03.2025) G-II, 3.6
2. EPO, IP5 Roundtable on AI: Summary Report (22.05.2019)
3. a.g.e. pg. 468.
4. Nari Lee, Protection for Artificial Intelligence in Personalised Medicine – The Patent-Trade Secret Tradeoff, The Harmonization and Protection of Trade Secrets in the EU (Edward Elgar Publishing 2020)
5. Recentive Analytics, Inc. v. Fox Corp., Fox Broadcasting Company, LLC, Fox Sports Productions, LLC, Case No. 23-2437, (Fed. Cir. 2025).
6. EPO, IP5 Roundtable on AI: Summary Report (22.05.2019)
7. EPO, IP5 Roundtable on AI: Summary Report (22.05.2019)
8. Tabrez Y. Ebrahim, Artificial Intelligence Inventions & Patent Disclosure, 125 Penn State Law Review 147 (2020).
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
